• Add STMicroelectronics: second iteration of the neural networking technology.
• Add Google's Pixel Visual Core.
• Add Intel Nervana™ Neural Network Processors (NNP).
• Add Intel's Loihi test chip.
• Add Nvidia open source Deep Learning Accelerator.
• Add Imagination reveals PowerVR Neural Network Accelerator (NNA). 
• Add Tesla said to be working on its own self-driving AI chip with AMD.
• Add Apple unveils A11 bionic neural engine AI chip. 

Myriad™ X is the first VPU to feature the Neural Compute Engine - a dedicated hardware accelerator for running on-device deep neural network applications. Interfacing directly with other key components via the intelligent memory fabric, the Neural Compute Engine is able to deliver industry leading performance per Watt without encountering common data flow bottlenecks encountered by other architectures.

Qualcomm has been fussing around ML for a while with the Zeroth SDK and Snapdragon Neural Processing Engine. The NPE certainly works reasonably well on the Hexagon DSP that Qualcomm use. The Hexagon DSP is far from a very wide parallel platform and it has been confirmed by Yann LeCun that Qualcomm and Facebook are working together on a better way in Wired's "The Race To Build An AI Chip For Everything Just Got Real", "And more recently, Qualcomm has started building chips specifically for executing neural networks, according to LeCun, who is familiar with Qualcomm's plans because Facebook is helping the chip maker develop technologies related to machine learning. Qualcomm vice president of technology Jeff Gehlhaar confirms the project. "We're very far along in our prototyping and development," he says." Perhaps we'll see something soon beyond the Kryo CPU, Adreno GPU, Hexagon DSP, and Hexagon Vector Extensions. It is going to be hard to be a start-up in this space if you're competing against Qualcomm's machine learning.

In a recent blog, "We are making on-device AI ubiquitous" shows its AI roadmap.

Nvidia's latest GPU can do 15 TFlops of SP or 120 TFlops with its new Tensor core architecture which is a FP16 multiply and FP32 accumulate or add to suit ML.

Nvidia is packing up 8 boards into their DGX-1for 960 Tensor TFlops.

Nvidia Volta - 架构看点 gives some insights of Volta architecture.

Now the open source DLA is available on Github and more information can be found here.

The NVIDIA Deep Learning Accelerator (NVDLA) is a free and open architecture that promotes a standard way to design deep learning inference accelerators. With its modular architecture, NVDLA is scalable, highly configurable, and designed to simplify integration and portability. The hardware supports a wide range of IoT devices. Delivered as an open source project under the NVIDIA Open NVDLA License, all of the software, hardware, and documentation will be available on GitHub. Contributions are welcome.

AMD has put a very good X86 server processor into the market for the first time in nine years, and it also has a matching GPU that gives its OEM and ODM partners a credible alternative for HPC and AI workload to the combination of Intel Xeons and Nvidia Teslas that dominate hybrid computing these days.

Tesla is reportedly developing its own processor for artificial intelligence, intended for use with its self-driving systems, in partnership with AMD. Tesla has an existing relationship with Nvidia, whose GPUs power its Autopilot system, but this new in-house chip reported by CNBC could potentially reduce its reliance on third-party AI processing hardware.

Xilinx provide "Machine Learning Inference Solutions from Edge to Cloud" and naturally claim their FPGA's are best for INT8 with one of their white papers.

Whilst performance per Watt is impressive for FPGAs, the vendors' larger chips have long had earth shatteringly high chip prices for the larger chips. Finding a balance between price and capability is the main challenge with the FPGAs.

ISSCC2017 Deep-Learning Processors文章学习 （一） is a reference.

Mobile Camera SoC
According to a Brief Data Sheet of Hi3559A V100ESultra-HD Mobile Camera SoC, it has:

Dual-core CNN@700 MHz neural network acceleration engine

Google's original TPU had a big lead over GPUs and helped power DeepMind's AlphaGo victory over Lee Sedol in a Go tournament. The original 700MHz TPU is described as having 95 TFlops for 8-bit calculations or 23 TFlops for 16-bit whilst drawing only 40W. This was much faster than GPUs on release but is now slower than Nvidia's V100, but not on a per W basis. The new TPU2 is referred to as a TPU device with four chips and can do around 180 TFlops. Each chip's performance has been doubled to 45 TFlops for 16-bits. You can see the gap to Nvidia's V100 is closing. You can't buy a TPU or TPU2. Google is making them available for use in their cloud with TPU pods containing 64 devices for up to 11.5 PetaFlops of performance.

Pixel Visual Core is Google’s first custom-designed co-processor for consumer products. It’s built into every Pixel 2, and in the coming months, we’ll turn it on through a software update to enable more applications to use Pixel 2’s camera for taking HDR+ quality pictures.

Other references are:
Google TPU 揭密

Google的神经网络处理器专利

脉动阵列 - 因Google TPU获得新生

Should We All Embrace Systolic Arrays?

PowerVR Series2NX Neural Network Accelerator

1. FPGAs and AI processors: DNN and CNN for all
   
2. 12 AI Hardware Startups Building New AI Chips
   
3. Tutorial on Hardware Architectures for Deep Neural Networks
   
4. Neural Network Accelerator Inference